Deep Learning Transformers and Attention 3

Question 1

What is prompt engineering in the context of large language models?

Accepted Answer

Crafting input text to elicit desired model behavior without changing model parameters

Answer

Modifying model weights to improve performance

Answer

Designing the tokenization scheme for the model

Answer

Engineering hardware prompts for GPU acceleration

Question 2

What is fine-tuning a pretrained Transformer on a downstream task?

Accepted Answer

Continuing training of pretrained weights on labeled task data to adapt representations to the target domain

Answer

Retraining the model from scratch on task-specific data

Answer

Using the pretrained model only as a feature extractor with a frozen backbone

Answer

Replacing attention heads with task-specific modules

Question 3

What is the attention mask used for in Transformer models during batched training?

Accepted Answer

Preventing attention to padding tokens to avoid polluting representations with pad token information

Answer

Masking gradient flow through specific attention heads

Answer

Forcing all attention weights to be equal

Answer

Preventing the model from attending to low-frequency tokens

Question 4

What is a token in the context of Transformer language models?

Accepted Answer

A unit of text (word, subword, or character) that the model processes as a single input element

Answer

A single character of text

Answer

A single row in the embedding matrix

Answer

A hyperparameter controlling vocabulary size

Question 5

What is RLHF (Reinforcement Learning from Human Feedback) used for in LLM training?

Accepted Answer

Aligning the model's outputs with human preferences by using human ratings as a reward signal

Answer

Pretraining the base language model on internet text

Answer

Improving the tokenizer's vocabulary using human-labeled text

Answer

Optimizing inference speed based on human latency preferences

Question 6

What does the term 'context window' refer to in a large language model?

Accepted Answer

The maximum number of tokens the model can process in a single forward pass

Answer

The model's training dataset size

Answer

The number of attention heads in the model

Answer

The sliding window used during data preprocessing